Generic Reading Recognition Method and System for Pointer-type Meters
There are various styles of pointer-type instruments,and existing methods can only perform conventional readings on instruments with one or two pointers,and cannot simultaneously support situations with more than two dials,more than two pointers,and multiple function meters.Based on the above problems,a general pointer-type instrument reading recognition method based on deep learning is proposed in this paper.Firstly,the YOLOv8 object detection method is used to construct a dial detection model for pointer-type instruments to locate the dial position.Then further construct a pointer segmentation model in the dial using Mask R-CNN instance segmentation method to identify the pointer mask re-gion.A calibration tool is built to calibrate the dial information and pass the parameters to the reading recognition module.At the same time,this paper builds a preset position correction algorithm to correct the preset position when it is offset,so as to ensure that the calibrated information and the dial position keep corresponding.A general reading recognition module for pointer-type meters is developed,which can traverse the case of multiple pointers and dials and realize the reading recognition of any type of pointer meters.